Built by Behavioral Robotics, our infrastructure arm
Every movement of the face, voice, and body, captured frame by frame and turned into data you can search, compare, and act on.
See what we captureThe Nonverbal Group · Capture
Every conversation we record is measured along hundreds of behavioral channels, frame by frame. Here are six of them.
Smiles.
Every smile measured from onset to offset: intensity, symmetry, timing, and how long it holds.
Prosody.
The melody of speech. Pitch, pace, volume, and the pauses between words, tracked across every turn.
Facial heatmaps.
Where the face moves, and how much, mapped region by region across the conversation.
Eyes.
Gaze, blinks, and eye movement. Where attention goes, when it breaks, and the rhythm underneath.
Linguistic analysis.
What gets said, and how. Word choice, hedges, and interruptions, aligned to the face and the voice.
Facial configurations.
How the face arranges itself, moment to moment: brows, eyes, mouth, and jaw measured together.
2009–2019
We started with good old-fashioned manual labeling. Hours of conversations watched frame by frame, every blink, expression and gesture tagged by a trained coder. Slow and painful, but it taught us exactly what to look for.
2019–2025
We automated with the best open tools available: generic face tracking plus AI review. It scaled the work, but a generic tool sees a generic face. It missed subtle signals, struggled with real-world footage, and couldn't show its work.
2025–2026
So we built our own. A 3D model of each person's face, fitted to every frame, with components trained on thousands of frames our experts labeled by hand. Every measurement is drawn on the video, so you can see that it's right.
2026–2028
Faster, sharper and richer. The system learns from every expert correction, runs at the speed of video, and covers more of the body and conversation.
2028 →
The endpoint: the same perception running in real time, as the behavioral layer for machines that interact with people.
From the Content Lab
The “93% of communication is nonverbal” rule is a misreading of two 1967 studies by Albert Mehrabian. Where the number came from, what he actually said, and three lessons about how research gets misrepresented.
